Bacterial Colonization of Early Stages of Limnetic Diatom Microaggregates

AQUATIC MICROBIAL ECOLOGY Vol. 25: 141–150, 2001 Published September 4 Aquat Microb Ecol

Bacterial colonization of early stages of limnetic diatom microaggregates

Sandra Knoll, Walter Zwisler, Meinhard Simon*

Institute for Chemistry and Biology of the Marine Environment, University of Oldenburg, 26111 Oldenburg, Germany

ABSTRACT: Macroscopic organic aggregates in pelagic environments are colonized by bacterial populations that differ from those in the surrounding water. To understand better how this well- adapted bacterial community is established, it is important to examine the initial colonization early in the aggregation process. We studied, therefore, the early formation and bacterial colonization of diatom microaggregates (MA) (<150 µm) during the phytoplankton spring bloom in Lake Constance, Germany. Water samples were incubated in plexiglass cylinders rolled horizontally for 44 to 48 h and subsampled for MA and their bacterial colonization, which was examined by fluorescent in situ hybridization with oligonucleotides of various specificity. During the initial 24 h bacterial numbers remained ~70 cells MA–1 and finally increased to ~250 MA–1. Detection rates of Bacteria by probe EUB338 ranged from 40% to >80% of the DAPI-stainable cells. Initially, α-Proteobacteria detected by the probe ALF1b dominated the bacterial community on MA, whereas toward the end of the incubation β-Proteobacteria increasingly dominated. Proportions of Cytophaga/Flavobacteria detected by the probe CF319a also increased systematically on MA toward the end but constituted lower proportions than β-Proteobacteria. In the surrounding water β-Proteobacteria dominated during the initial 24 h whereas Cytophaga/Flavobacteria consistently dominated in the late phase of the experiments. Applying highly specific probes for narrow clusters of close relatives of the genus Sphingomonas (α-Proteobacterium), Duganella zoogloeoides (formerly Zoogloea ramigera) and Aci- dovorax facilis (both β-Proteobacteria), we found that these bacteria were present on MA already at the initial sampling or at the latest after 10 h and comprised substantial and sometimes dominant proportions of total α- and β-Proteobacteria. These bacteria, also dominant on natural lake snow aggregates in Lake Constance, were never detected among the free-living bacterial community in the surrounding water. Hence, our results indicate that the bacterial community on lake snow aggregates develops largely from seeds on their precursor MA.

KEY WORDS: Aggregates · Diatoms · Bacteria · In situ hybridization · Lake Constance

Resale or republication not permitted without written consent of the publisher

INTRODUCTION either attached to algal cells or released as transparent exopolymer particles (TEP), have been identified as a The formation of microscopic (<500 µm; micro- key agent leading to the aggregation event and thus to aggregates [MA]) and macroscopic organic aggre- the formation of a unique microhabitat highly enriched gates (>500 µm; marine snow, lake snow) during the in nutrients, protozoans and bacteria compared with course and breakdown of diatom blooms is a well- the dilute bulk environment (Shanks & Trent 1979, known phenomenon, eventually leading to their sedi- Caron et al. 1982, Alldredge & Silver 1988, Alldredge mentation (Smetacek 1985, Alldredge & Gotschalk et al. 1993, Passow et al. 1994, Logan et al. 1995, 1989, Kiørboe et al. 1994). Acidic polysaccharides, Grossart & Simon 1998). Specific populations of bacteria attached to phytoplankton cells or in the phycos- phere lead to the formation of different types of *Corresponding author. E-mail: [email protected] exopolysaccharides by diatoms (Grossart 1999). It is,

however, still unclear whether the bacterial coloniza- composition of the bacterial community on aggregates tion is instrumental in the formation of aggregates. and that in the surrounding water, i.e., how the free- There is good evidence, though, that the aggregate- living bacterial community interacts with that on MA associated heterotrophic bacterial community rapidly and macroaggregates. respires and solubilizes at least the labile aggregate We studied the bacterial colonization and composi- components (Smith et al. 1992, Grossart & Simon 1998, tion of the bacterial community on early stages of Berman et al. 1999, Ploug et al. 1999, Grossart & Ploug diatom MA, formed in rolling tanks according to 2000). Shanks & Edmondson (1989) during the phytoplankton The composition of the bacterial community associ- spring bloom in Lake Constance. By applying rRNA- ated with macroscopic aggregates (macroaggregates) targeted fluorescent oligonucleotides of various speci- is highly specific and differs from that of the free-living ficity, and including highly specific ones for narrow bacterial community in the surrounding water. Quali- clusters of α- and β-Proteobacteria, we found that the tative differences between bacterial communities asso- bacteria dominating during later stages on aggregates ciated with marine snow and those in the surrounding were already present on the initial stages in substantial water have been found in various marine environ- amounts. The aggregate-specific bacteria detected by ments (DeLong et al. 1993, Rath et al. 1998, Acinas et the highly specific probes, however, were never al. 1999, Phillips et al. 1999, Simon et al. in press). detected in the surrounding water. Highly specific bacterial communities have also been found on lake snow aggregates (Weiss et al. 1996, Schweitzer et al. 2001) and riverine aggregates (Böck- MATERIALS AND METHODS elmann et al. 2000). By applying fluorescent in situ hybridization with a few highly specific 16S rRNA Samples for the experiments were collected on 28 oligonucleotide probes, targeting narrow clusters of April (Expt 1), and 5 May (Expt 2) and 12 May (Expt 3) β1-subclass-Proteobacteria and of α-Proteobacteria 1998 at 3 m depth at the center of Lake Überlingen, the closely related to the genus Sphingomonas, Schweitzer northwestern arm of mesotrophic Lake Constance, et al. (2001) identified 16 to 53% of the DAPI-stainable Germany. Lake Constance is a warm monomictic pre- cells on lake snow aggregates in Lake Constance, Ger- alpine lake with a surface area of 472 km2, and maxi- many. In the course of a few days and while the lake mum and mean depths of 253 and 101 m, respectively. snow aggregates sank through the water column, the The lake has been studied intensively in the recent composition of their bacterial community changed past (Gaedke 1998, Häse et al. 1998, Simon et al. 1998), from a slight dominance of α-Proteobacteria to a pro- also with respect to the significance of lake snow nounced dominance of β-Proteobacteria. This transi- aggregates (Grossart & Simon 1993, 1998, Weiss et al. tion was paralleled by that of a consumption of dis- 1996, Grossart et al. 1997, Schweitzer et al. 2001). solved amino acids from the surrounding water to their Because the age of natural aggregates cannot be release into it. Interestingly, the composition of the correctly determined, we applied an experimental lake snow-associated bacterial community differed approach using rolling tanks according to Shanks & from that of MA collected during the same period in Edmondson (1989). To examine the bacterial coloniza- the lake (Brachvogel et al. 2001). α-Proteobacteria tion of truly early stages of aggregates we studied the were not detected on these MA whereas cells of the formation of MA (<150 µm) of the freshly collected Cytophaga/Flavobacteria cluster were relatively more samples and their bacterial colonization in these tanks abundant than on lake snow aggregates. over a period of 44 to 48 h. Within 2 h of collection, the To understand better the colonization and establish- samples were transferred into a series of sterilized 1.2 l ment of specific bacterial communities on macroaggre- plexiglass cylinders and further incubated rolling hori- gates, the very early colonization stages and the initial zontally at 2 rpm at in situ temperature and irradiated conditions need to be compared with those in the sur- with 200 µE m–2 s–1 in a 12:12 h light:dark cycle. This rounding water. Azam et al. (1993) studied the initial irradiance is the mean of the growing season at 3 m colonization and growth of bacteria on freshly formed depth in Lake Constance. During the incubation the diatom MA and found the highest community growth tanks were sampled periodically for MA and their bac- rates during the earliest stages. To understand the col- terial colonization. The surrounding water was sam- onization process better it is important to know pled as well at each time point. Since each tank was whether the bacteria dominating during later stages on sampled only once, time series were composed of sam- aggregates are already present on very early stages ples from different tanks. Earlier experiments had with a low bacterial colonization or whether they shown that the formation of aggregates and their appear only later. A related question is whether during bacterial colonization in tanks incubated in parallel these early stages any relation exists between the were not statistically different (Schweitzer et al. 2001). Knoll et al.: Bacteria on diatom aggregates 143

For examination of the MA and the associated bacter- Table 1. Total numbers of bacteria (±SD, n = 3) per micro- ial community 20 ml was filtered onto 12 µm pore size aggregate (MA) and in the surrounding water during Expt 1, Nuclepore filters (47 mm diameter, Costar Corp., Cam- Expt 2 and Expt 3, carried out on 28 April, and 5 and 12 May 1998 bridge, MA, USA) by gravity. For examination of the free-living bacterial community 2 ml was filtered onto 0.2 µm pore size Nuclepore filters (25 mm diameter) Time Microaggregates Surrounding water (h) (cells MA–1)(106 cells ml–1) with a gentle vacuum of 150 mbar. Five samples were filtered at each time point such that triplicate samples Expt 1 1 of ⁄4 of a filter were available for in situ hybridization of 3 71 ±32 1.1 ±0.1 each probe (see below). Fixation of the bacteria was 15 69 ±32 1.3 ±0.2 24 68 ±29 1.1 ±0.2 done by overlaying the filter with 4% paraformalde- 41 101 ± 530 1.5 ± 0.1 hyde-PBS (pH 7.2) for 30 min. After withdrawing the 48 254 ± 131 1.8 ± 0.2 fixative, the filters were rinsed with 2 ml of PBS and Expt 2 Milli-Q water. The samples were stored at 4°C until 2 64 ±23 1.1 ±0.7 further processing within a few weeks. Numbers and 14 66 ±24 1.5 ±0.1 sizes of MA and total numbers of bacteria on aggre- 24 74 ±24 1.6 ±0.2 30 117 ± 105 2.0 ± 0.3 gates and in the surrounding water were also deter- 45 267 ± 193 2.9 ± 0.4 mined by epifluorescence microscopy in DAPI-stained Expt 3 samples of the volumes indicated above and filtered 2 68 ±40 1.3 ±0.3 onto 12 and 0.2 µm Nuclepore filters, respectively 13 60 ±29 1.9 ±0.1 (Porter & Feig 1980). The abundance of MA was deter- 24 87 ±37 3.2 ±0.4 mined at 400× magnification on 10 viewfields. Thirty 37 84 ±30 2.5 ±0.5 45 265 ± 181 3.1 ± 0.7 MA were sized according to their longest dimension. For bacteria on aggregates, DAPI-stainable and probe- specific bacteria were counted on 20 randomly selected MA. Because of the variable size of the MA, (Huber 1997). The target molecule of probes BET42a the coefficient of variation (CV; SD/mean) of the DAPI- and GAM42a is the 23S rRNA and that of all other cell counts ranged from 20 to 50% and in some cases probes the 16S rRNA. The probes were linked to the up to 90% (Table 1, Figs 1 & 2). In samples of the sur- fluorochrome 5(6)-carboxy-fluorescein-N-hydroxysuc- rounding water 400 to 500 cells were counted on 10 cinimide ester (FLUOS) or Cy3 (derivate of succinimide viewfields. In these samples the CV was always <15%. ester Cy3 of a cyanine). For the analysis of the aggre- In situ hybridization. The composition of the bacter- gate-associated bacterial community, the probes were ial community was determined by in situ hybridization linked to FLUOS, except the CF319a probe, which was with the fluorescence-labeled rRNA targeted oligonu- always linked to Cy3. Free-living bacteria were ana- cleotides listed in Table 2. The probes LSA225, LSB65 lyzed exclusively with Cy3-labeled probes because and LSB145 were designed from lake snow clone li- they have a higher intensity than FLUOS. Because of braries of Lake Constance in which the clone clusters the smaller size of the free-living bacteria (<0.08 µm3) targeted by these probes were prominent members than that of the aggregate-associated cells (0.1 to

Fig. 1. Bacterial colonization of microaggregates in Expt 1 from 28 April 1998. Total numbers of cells and of Bacteria detected by (A) probe EUB338; (B) numbers; and (C) percentages of DAPI-stainable cells of α- (ALF) , β- (BET) and γ-Proteobacteria (GAM), and of the Cytophaga/Flavobacteria cluster (CF) 144 Aquat Microb Ecol 25: 141–150, 2001

Fig. 2. Numbers of bacteria in the surrounding water in (A–C) Expt 1 (28 April) and (D–F) Expt 3 (12 May). Numbers of total cells and of Bacteria detected by (A,D) probe EUB338; (B,E) numbers; and (C,F) percentages of DAPI-stainable cells of α-, β- and γ-Proteobacteria, and of the Cytophaga/Flavobacteria cluster. See Fig. 1 for abbreviations

0.2 µm3), this differential labeling maximized the frac- oligonucleotide (Manz et al. 1992, Snaidr et al. 1997). tion of free-living bacteria that was detected but did not To stop hybridization, the filter on the glass slides were affect that of the aggregate-associated cells. in situ hy- rinsed and incubated at 46°C for 15 min in washing bridization was performed at 46°C for 90 min in pre- buffer, containing 180 mM (20%) or 40 mM (35%) heated hybridization chambers. The procedures of formamide, NaCl and 0.01% SDS. After rinsing with Manz et al. (1992, 1996) were applied but modified as Milli-Q water, the slides were dried and stained with follows. Each filter was cut into quarters, of which 2 DAPI (0.01%). Finally, they were embedded in City- were put on 1 glass slide for in situ hybridization. To fluor (Chemical Laboratory, University of Kent, Kent, each filter quarter, 20 µl of the hybridization buffer was UK) on glass slides and covered with cover slides. The applied. It contained 0.9 M NaCl and formamide at a hybridized samples were visualized by a Nikon epifluo- concentration of 20% (EUB338, ALF1b) or 35% (other rescence microscope equipped with the filter sets UV- probes), 20 mM Tris-HCl (pH 7.4), 0.01% SDS and the 2A for DAPI (Nikon), B-2A for FLUOS (Nikon, Tokyo, probe in a concentration of 2.5 ng µl–1 each. Probes Japan) and XF32 NM198 for Cy3 (Omega Optical Inc., BET42a and GAM42a were used with a competitor Brattleboro, VT, USA) at a magnification of 1250×. As

Table 2. Identification, target groups and sequences according to Brosius et al. (1981) of the oligonucleotide probes used. For the probes specific for α- (LSA) and β-subclass Proteobacteria (LSB), the closest related and identified bacterial species is given

Probe Target group Sequence (5’→3’) Source

EUB33 Bacteria GCT GCC TCC CGT AGG AGT Amann et al. (1990) ALF1b α-Proteobacteria CGT TCG (C/T)TC TGA GCC AG Manz et al. (1992) BET42a β-Proteobacteria GCC TTC CCA CTT CGT TT Manz et al. (1992) GAM42a γ-Proteobacteria GCC TTC CCA CAT CGT TT Manz et al. (1992) CF319a Cytophaga/Flavobacteria TGG TCC GTG TCT CAG TAC Manz et al. (1996) LSA225 Sphingomonas sp. and relatives TCC TAC GCG GGC TCG TCC Schweitzer et al. (2001) LSB65 Duganella zoogloeoides and relatives GTT GCC CCG CGC TGC CGT Schweitzer et al. (2001) LSB145 Acidovorax facilis and relatives CTT TCG CTC CGT TAT CCC Schweitzer et al. (2001) Knoll et al.: Bacteria on diatom aggregates 145

negative controls for the highly specific probes we used gluing the algal cells together. During the initial hours Agrobacterium rubi (LSA225), Ectothiorhodospira of the experiments, together with the diatoms, MA of a halophila (LSB65) and Hydrogenophaga palleroni size of 50 to 80 µm were present at abundances of (LSB145). Cells were counted as described above. The <50 ml–1. After 24 to 30 h the mean size of the MA CV of the triplicate analyses of the various probes increased to >100 µm and the abundance to 60 to ranged between 0.06–0.34 (EUB338), 0.12–0.49 80 ml–1 even though MA of the initial smaller size class (ALF1b), 0.12–0.49 (BET42a), 0.10–2.20 (GAM 42a), were still present. When the experiments were and 0.11–1.50 (CF319a). The CV exceeded 0.6 when stopped after 44 to 48 h, a few aggregates of >3 mm very low cell numbers were recorded. had formed in each tank. Total numbers of bacteria on the MA subsampled 2 to 3 h after the start of the experiments ranged from RESULTS 64 to 80 cells MA–1 and remained constant within 24 h (Table 1, Fig. 1A). Thereafter the numbers of MA- Samples were collected at the height and break- associated bacteria increased to 254 to 267 cells MA–1 down of the phytoplankton spring bloom as indicated at the end of the experiments. Bacterial numbers in the by chlorophyll a concentrations at 3 m of 8.3, 12.6 and surrounding water initially were ∼1 × 106 ml–1 and 2.3 µg l–1 on 28 April, and 5 and 12 May, respectively increased to 1.8 × 106 ml–1 in Expt 1, and to 3 × 106 ml–1 (B. Beese unpubl. data). The microscopic examination in Expts 2 and 3 (Table 1, Fig. 2A,D). Because the of the DAPI-stained samples withdrawn from the results of the bacterial colonization of the MA in the 3 rolling tanks showed that the phytoplankton was dom- experiments were rather similar, we present the inated by diatoms such as Asterionella formosa, Fragi- detailed results only of Expt 1. laria crotonensis and Stephanodiscus spp. MA were Probe EUB338 detected at least 40% of the DAPI- always composites of diatoms and organic material stainable cells on MA. Toward the end of the experi-

Fig. 3. Colonization of microaggregates by specific bacterial populations during Expt 1 (28 April), Expt 2 (5 May) and Expt 3 (12 May). Left panels: percentages of α-Proteobacteria detected by the probe LSA225 (Sphingomonas and relatives); central panels: percentages of β-Proteobacteria detected by the probes LSB145 (Duganella zoogoeoides and relatives) and LSB65 (Acidovorax facilis and relatives); right panels: cumulative percentages of DAPI-stainable bacteria detected by the probes LSA225, LSB145 and LSB65 146 Aquat Microb Ecol 25: 141–150, 2001

ments the detection rate increased to >70%. The were present already at the initial sampling at 2 h and detection rate of the free-living bacteria by probe comprised proportions of 12% of the DAPI cell counts EUB338 was also 40 to 80% of the DAPI-stainable cells during the entire experiment (Table 3). All cells of the but only 20 to 40% in Expt 3 (Figs 2D & 3A). α- Proteobacteria subclasses and Cytophaga/Flavobacte- Proteobacteria comprised 25 to 45% of the DAPI- ria together accounted for ∼100% of the counts by stainable cells on MA in all 3 experiments with only probe EUB338. minor fluctuations during a single experiment The composition of the free-living bacterial commu- (Fig. 1C). Proportions during the initial 24 h were not nity in the surrounding water differed greatly from significantly different from those during the late period that on MA. During the initial 24 h it was dominated after 30 to 48 h (t-test, p < 0.01; Table 3). The propor- largely by β-Proteobacteria comprising 15 to 25% of tions of β-Proteobacteria on MA during the first 24 h the DAPI cell counts. These proportions were always were lower than those of α-Proteobacteria but signifi- significantly higher than those of α-Proteobacteria cantly increased toward the end of the experiments. In (Table 3), which comprised <10% and often <5% the late phase of Expt 1 and Expt 2 they were signifi- (Fig. 3B,C,E,F). In contrast to the aggregate-associ- cantly higher than those of α-Proteobacteria (t-test, p < ated β-Proteobacteria, proportions of this phyloge- 0.01) and in Expt 3 were not significantly different netic cluster in the surrounding water remained con- from those of α-Proteobacteria (Table 3). Proportions of stant (Expt 1) or decreased in the late phase (Expt 2 γ-Proteobacteria on MA always comprised substan- and Expt 3). An inverse trend was also true for α-Pro- tially lower proportions than α- and β-Proteobacteria teobacteria, whose proportions increased in 2 of the 3 but also consistently increased during the course of the experiments during the late phase. γ-Proteobacteria experiments (Fig. 1B,C, Table 3). Cells of the Cyto- always comprised only minor proportions. Even phaga/Flavobacteria cluster in Expt 1 and Expt 2 though cells of the Cytophaga/Flavobacteria cluster occurred only occasionally during the initial 24 h but occurred in the surrounding water at the same time as accounted for 9.7 and 15.8% of the DAPI-stainable on MA, their proportions in the former were signifi- cells, respectively, in the late phase (Fig. 1B,C, cantly higher and accounted for at least 16% but after Table 3). In Expt 3, cells of this phylogenetic cluster 24 h for 24 to 41% of the DAPI-stainable cells (Fig. 2B,C,E,F). During the late phase they were the single most abundant bacterial group. In Table 3. Proportions of α- (ALF), β- (BET) and γ-Proteobacteria (GAM) and of Cytophaga/Flavobacteria (CF) on microaggregates Expt 1 and Expt 2 the free-living bacteria (MA) and in the surrounding water (bulk) during the initial and late detected by the group-specific probes ac- phase of the Expt 1 (28 April), Expt 2 (5 May) and Expt 3 (12 May). counted for 50% of the EUB338 counts during Mean values ± SE for the respective periods are given the initial 24 h but for 100% thereafter. In Expt 3 they accounted for 100% of the EUB338 Time ALF BET GAM CF counts during the entire course. (h) (%DAPI) (%DAPI) (%DAPI) (%DAPI) Bacteria related closely to Sphingomonas sp. detected by probe LSA225 were present already Expt 1 MA at the initial sampling in Expt 1 and Expt 2 and, 3–24 33.5 ± 3.0 21.5 ± 2.5 5.0 ± 0.7 0.3 ± 0.7 after 12 h and including Expt 3, comprised pro- 41–48 30.6 ± 5.3 34.5 ± 0.2 12.6 ± 1.6 9.7 ± 1.2 portions of 10 to 32% of total α-Proteobacteria bulk (Fig. 3, left panels). Duganella zoogloeoides (for- 3–24 13.9 ± 0.2 14.9 ± 0.8 3.1 ± 0.3 0.4 ± 0.4 merly Zoogloea ramigera ATCC 25935) and rel- 41–48 10.2 ± 1.4 17.9 ± 0.6 4.5 ± 0.3 40.7 ± 2.1 atives and Acidovorax facilis and relatives, Expt 2 detected by probes LSB65 and LSB145, were MA 2–24 27.5 ± 1.6 25.5 ± 2.6 0.6 ± 0.8 0 also detected already at the initial sampling 30–44 25.5 ± 1.4 29.5 ± 1.0 5.6 ± 1.5 15.8 ± 0.5 (Fig. 2, center panels). Probe LSB65 in most Bulk cases detected as much as 27 to 70% of total β- 2–24 4.0 ± 0.4 20.7 ± 0.9 0.6 ± 0.2 0 Proteobacteria and always higher proportions 30–44 3.5 ± 0.5 14.8 ± 1.2 1.3 ± 0.1 35.9 ± 1.9 than probe LSB145. Probes LSB65+LSB145 Expt 3 together in some cases detected 60 to 100% of MA total β-Proteobacteria. All 3 probes together 2–24 38.7 ± 2.7 15.8 ± 2,3 9.0 ± 1.4 11.5 ± 1.5 detected between 2 and 28% of the DAPI-stain- 37–45 38.0 ± 1.8 35.3 ± 1.0 12.7 ± 1.2 12.0 ± 0.3 able bacteria (Fig. 3, right panels). None of the 3 Bulk 2–24 18.2 ± 0.1 18.9 ± 4.2 2.4 ± 0.1 20.8 ± 1.2 highly specific probes detected any bacteria 37–45 19.1 ± 0.8 13.9 ± 0.5 1.5 ± 0.2 31.3 ± 1.8 among the free-living bacterial community in the surrounding water. Knoll et al.: Bacteria on diatom aggregates 147

DISCUSSION with observations made by Azam et al. (1993) during the aggregation of marine diatoms. Our experiments with lake water incubated in In our experiments the bacterial numbers MA–1 rolling tanks showed that MA of a size of 50 to 80 µm ranged between 61 and 80 for the initial 24 h when the increased to >100 µm and formed macroaggregates size of the MA was still <80 µm but increased there- within 40 h. This time period of macroaggregate for- after with doubling times of 5 to 12 h until 44 to 48 h. mation is very similar to that reported previously for This increase was presumably due to cell multiplica- the same lake (Weiss et al. 1996, Grossart et al. 1997, tion as well as gluing together of smaller MA. Similar Schweitzer et al. 2001). Many studies in various envi- observations with even higher growth rates are ronments and including Lake Constance have shown reported by Azam et al. (1993). Our bacterial numbers that TEP, either as single MA or associated with MA–1 are well in the range of other reports including diatom cells, are of key importance in gluing together TEP, protein containing MA, and DAPI yellow particles these phytoplankton algae and forming macroaggre- and other types of MA (Passow et al. 1994, Mostajir et gates (Alldredge et al. 1993, Kiørboe et al. 1994, Pas- al. 1995, Long & Azam 1996, Brachvogel et al. 2001). sow & Alldredge 1994, Passow et al. 1994, Logan et al. Bacteria of close relatives of Sphingomonas, 1995, Grossart et al. 1997). Therefore, even though Duganella zoogloeoides and Acidovorax facilis de- TEP was not examined specifically, we assume that it tected by probes LSA225, LBS 65 and LSB 145, respec- primarily mediated the aggregation in our experi- tively, were present on MA already at the initial sam- ments. Small MA were still present after 40 h, pre- pling or after 10 h of incubation (LSA225) and sumably because the phytoplankton was still growing continuously constituted substantial and sometimes in the rolling tanks incubated in a light-dark cycle dominating proportions of total α- and β-Proteobacte- and continuously produced new MA. We did not ria. In the 3 experiments detection rates of Bacteria and examine the formation of TEP from dissolved precur- of total group-specific cells initially were 40%, but in sor material but did examine the formation of MA of most cases 55 to 75%, of the DAPI-stainable cells and diatom origin, which were always identified by the increased to >80%, indicating that the majority of the diatom structures. The MA were colonized by a bac- bacterial community on MA was detected by the terial community dominated by narrow clusters of α- probes applied. The initial detection rates are at the and β-Proteobacteria and very similar to those on nat- upper end of those of natural MA in Lake Constance ural and naturally derived lake snow aggregates sev- and of reports of free-living bacterial communities in eral days old in Lake Constance (Schweitzer et al. lacustrine ecosystems (Glöckner et al. 1999, Pernthaler 2001). These authors examined the temporal succes- et al. 1999, Zwisler 2000, Brachvogel et al. 2001). In sion of the bacterial community on lake snow aggre- particular, bacteria detected by probe LSB65 and gates 2 to 4 d old and found a systematic decrease in closely related to D. zoogloeoides (formerly Zoogloea the proportion of α-Proteobacteria and an increase in ramigera) dominated total β-Proteobacteria to a great that of β-Proteobacteria such that the latter usually extent. These observations suggest that these bacteria exceeded the proportion of the former on lake snow are already present on senescent diatoms and are aggregates 3 to 4 d old. Hence, this succession is a actively involved in the aggregation process. D. consistent follow-up of the one we found and indi- zoogloeoides is known to produce mucopolysac- cates that the MA we studied were indeed precursors charides, occurs on activated sludge flocs in sewage of lake snow aggregates. Natural MA <60 µm in Lake treatment plants and has also been found on mucilage Constance, stained by DAPI, were shown not to be of filamentous cyanobacteria (Caldwell & Caldwell precursors of lake snow aggregates but the older 1978, Ikeda et al. 1982, Dugan et al. 1992). α- and β- decomposition stages of lake snow aggregates and of Proteobacteria also dominated on natural lake snow other detrital particulate organic matter (Brachvogel aggregates in Lake Constance collected at 25 m depth et al. 2001). These authors showed that the abun- and on lake snow aggregates 2 to 3 d old and formed in dance of natural DAPI-stained MA did not covary rolling tanks (Schweitzer et al. 2001). On natural lake with that of phytoplankton biomass and TEP, and that snow aggregates, however, the proportions of β-Pro- their associated bacterial community was dominated teobacteria and of bacteria detected by the probe by β-Proteobacteria and cells of the Cytophaga/ LSB145 and closely related to A. facilis and relatives Flavobacteria cluster, whereas α-Proteobacteria were were higher than on MA, whereas that of bacteria never detected. The MA we studied presumably have detected by the probe LSB65 was lower. On the other a very short half-life under natural conditions and hand, the proportion of bacteria detected by the probe rapidly form or are scavenged by larger aggregates LSA225 and closely related to Sphingomonas sp. and because of their high sticking coefficient. Our experi- relatives was higher on lake snow aggregates than on mental observations and this conclusion are consistent MA, even though the proportion of total α-Proteo- 148 Aquat Microb Ecol 25: 141–150, 2001

bacteria was lower. As outlined in more detail by only in spring on riverine aggregates in the Elbe River, Schweitzer et al. (2001), aerobic chemoorganohetero- which in general were also dominated by β-Proteo- trophic bacteria occurring on activated sludge flocs are bacteria and specifically by populations closely related the closest known relatives of those identified on to Aquabacterium commune, which is not closely aggregates. These bacteria exhibit high potentials of related to the bacteria we detected on MA. hydrolytic enzyme activities and metabolize labile In the free-living bacterial community initial detec- organic matter such as amino acids and carbohydrates. tion rates of Bacteria in the 3 experiments ranged from Schweitzer et al. (2001) further showed that the bacte- 38 to 50% of the DAPI-stainable cells and that of total rial community on lake snow aggregates consumed group-specific bacteria from 22 to 44% with increasing labile organic matter, mainly amino acids, from the proportions during the course of the bloom and cover- surrounding water as long as it was dominated by α- ing 100% of Bacteria. At the end of the experiments, Proteobacteria, until an aggregate age of 2 to 3 d. proportions of 50 to 75% were detected. These detec- Thereafter, when β-Proteobacteria dominated more tion rates are well in the range of those found in free- and more, dissolved amino acids were released into living bacterial communities in other lacustrine ecosys- the surrounding water. Hence, we assume that during tems including Lake Constance (Glöckner et al. 1999, the early formation of aggregates we examined and Pernthaler et al. 1999, Zwisler 2000). In contrast to the when α-Proteobacteria dominated or shared equal bacterial community on MA, that in the surrounding proportions with β-Proteobacteria labile organic mat- water was not initially dominated by α-Proteobacteria ter was consumed from the surrounding water, pre- but by β-Proteobacteria. In the last experiment, the ini- sumably leading at least partly to the formation of tial proportion of bacteria of the Cytophaga/Flavobac- mucus-like sticky organic material and the aggrega- teria cluster was only slightly lower than that of β- tion of senescent diatom cells. Proteobacteria. During the course of the experiments Bacteria of the Cytophaga/Flavobacteria cluster mainly Cytophaga/Flavobacteria were growing such were also present on the MA we studied but consti- that at the end of all 3 experiments they clearly domi- tuted much lower proportions than α- and β-Pro- nated the bacterial community and constituted greater teobacteria. During the first 2 experiments they proportions than on MA. Hence, and in accordance occurred only after 40 and 30 h, respectively. In the last with the above mentioned assumption, the composi- experiment carried out at the decline of the phyto- tion of the bacterial community in the surrounding plankton spring bloom, however, they were already water suggests that during the experiments and the present at the initial time point. Bacteria of the spring bloom the supply and turnover of labile dis- Cytophaga/Flavobacteria cluster are known to de- solved organic matter decreased at the expense of grade a variety of complex polymers including recalci- polymeric and more refractory constituents. Interest- trant carbohydrates (Reichenbach 1992). Hence, we ingly, the free-living bacterial community in Lake Con- assume that the occurrence of bacteria of the Cyto- stance usually is dominated by β-Proteobacteria, and phaga/Flavobacteria cluster reflects the presence or Cytophaga/Flavobacteria constitute only about 50% of production on MA of low but increasing amounts of the former (Zwisler 2000). These findings indicate that more refractory organic matter during the course of the substrate supply to the free-living bacteria in our bloom. On natural MA in Lake Constance bacteria of experiments was somehow different from the long- the Cytophaga/Flavobacteria cluster constituted 8 to term in situ conditions and may reflect short-term 20% of the DAPI-stainable cells, equivalent to 23 to dynamics that become undetectable after a few days 37% of Bacteria detected by the probe EUB338 and when lower, i.e. weekly, sampling frequencies are (Brachvogel et al. 2001). This proportion was equal to applied. or only slightly lower than that of β-Proteobacteria, None of the probes highly specific for narrow clus- suggesting that natural MA in Lake Constance were ters of MA-associated α- and β-Proteobacteria de- much more refractory than the MA and precursors of tected any single free-living bacterium. This observa- macroaggregates we studied. On lake snow aggre- tion suggests that there was no measurable release gates in Lake Constance, however, bacteria of the and propagation of these aggregate-associated bacte- Cytophaga/Flavobacteria cluster were relatively less ria into the surrounding water. Obviously they were abundant than on natural MA and increased in propor- very closely associated with or even firmly attached to tion only at 50 m and below. Natural biofilm communi- the MA such that after division newly formed cells did ties in the Elbe River, Germany, and its estuary are also not reach the surrounding water. Hence, in the case of colonized by bacteria of this cluster to high proportions these bacteria, aggregates did not serve as a means of (Brümmer et al. 2000, Simon et al. in press). Böckel- their propagation into the surrounding water as mann et al. (2000), however, found high proportions of previously postulated (Jacobsen & Azam 1984). We this cluster (i.e. 30 to 40%) of DAPI-stainable bacteria hypothesize, however, that during later stages, when Knoll et al.: Bacteria on diatom aggregates 149

macroaggregates progressively decompose and disin- zation and primary structure of a ribosomal operon from tegrate, aggregate-specific bacteria are released and Escherichia coli. J Mol Biol 148:107–127 colonize newly produced senescent diatoms as seeds Brümmer IHM, Fehr W, Wagner-Döbler I (2000) Biofilm community structure in polluted rivers: abundance of domi- and may propagate again. Accordingly, they would nant phylogenetic groups over a complete annual cycle. propagate only a little in the substrate-dilute bulk Appl Environ Microbiol 66:3078–3082 water. Caldwell DE, Caldwell SJ (1978) Zoogloea ramigera sp. asso- In conclusion, we have shown that the bacterial ciated with blooms of Anabaena flos-aquae. Can J Micro- biol 24:922–931 community on diatom macroaggregates is already Caron DA, Davis PG, Madin LP, Sieburth JM (1982) Hetero- seeded and largely determined on their precursor MA trophic bacteria and bacterivorous protozoans in oceanic which obviously already provide a suitable microenvi- macroaggregates. Science 218:795–797 ronment. During the course of roughly 2 d the compo- DeLong EF, Franks DG, Alldredge AL (1993) Phylogenetic sition undergoes slight changes with increasing diversity of aggregate-attached vs. free-living marine bac- β terial assemblages. Limnol Oceanogr 38:924–934 proportions of -Proteobacteria and bacteria of the Dugan P, Stoner DL, Pickrum HM (1992) The genus Zoogloea. Cytophaga/Flavobacteria cluster. The significantly dif- In: Balows A, Trüper HG, Dworkin M, Harder W, Schleifer ferent composition of the MA-associated bacterial KH (eds) The prokaryotes. Springer, New York, community from that in the surrounding water, which p 3952–3964 Gaedke U (1998) Functional and taxonomical properties of exhibits much higher proportions of Cytophaga/ the phytoplankton community of large and deep Lake Flavobacteria, indicates pronounced differences in the Constance: interannual variability and response to re- turnover of organic matter between these 2 habitats. oligotrophication (1979–1993). Arch Hydrobiol Spec Iss Adv Limnol 53:119–141 Glöckner FO, Fuchs BM, Amann R (1999) Bacterioplankton Acknowledgements. We are grateful to B. Beese for making composition of lakes and oceans: a first comparison based available unpublished chlorophyll data. This work was sup- on fluorescence in situ hybridization. Appl Environ Micro- ported by grant Si 360/4-3 from the Deutsche Forschungs- biol 65:3721–3726 gemeinschaft. Grossart HP (1999) Interactions between marine bacteria and axenic diatoms (Cylindrotheca fusiformis, Nitzschia lae- vis, and Thalassiosira weissflogii) incubated under various LITERATURE CITED conditions in the lab. Aquat Microb Ecol 19:1–11 Grossart HP, Ploug H (2000) Bacterial production and growth Acinas SG, Antón J, Rodríguez-Valera F (1999) Diversity of efficiencies: direct measurements on riverine aggregates. free-living and attached bacteria in offshore western Limnol Oceanogr 45:436–445 Mediterranean waters as depicted by analysis of genes Grossart HP, Simon M (1993) Limnetic macroscopic organic encoding 16S rRNA. Appl Environ Microbiol 65:514–522 aggregates (lake snow): abundance, characteristics, and Alldredge AL, Gotschalk C (1989) Direct observation of the bacterial dynamics in Lake Constance. Limnol Oceanogr mass flocculation of diatom blooms: characteristics, set- 38:532–546 tling velocities, and formation of diatom aggregates. Grossart HP, Simon M (1998) Bacterial colonization and Deep-Sea Res 36:159–171 microbial decomposition of limnetic organic aggregates Alldredge AL, Silver ML (1988) Characteristics, dynamics and (lake snow). Aquat Microb Ecol 15:127–140 significance of marine snow. Prog Oceanogr 20:41–82 Grossart HP, Simon M, Logan BE (1997) Formation of macro- Alldredge AL, Passow U, Logan BE (1993) The abundance scopic organic aggregates (lake snow) in a large lake: the and significance of a class of large transparent organic significance of transparent exopolymer particles (TEP), particles in the ocean. Deep-Sea Res 40:1131–1140 phyto- and zooplankton. Limnol Oceanogr 42:1651–1659 Amann RI, Krumholz L, Stahl DA (1990) Fluorescent oligonu- Häse C, Gaedke U, Seifried A, Beese B, Tilzer MM (1998) cleotide probing of whole cells for determinative, phyloge- Phytoplankton response to re-oligotrophication in large netic, and environmental studies in microbiology. J Bacte- and deep Lake Constance: photosynthetic rates and riol 172:762–770 chlorophyll concentrations. Arch Hydrobiol Spec Iss Adv Azam F, Martinez J, Smith DC (1993) Bacteria-organic matter Limnol 53:159–178 coupling on marine aggregates. In: Guerrero R, Pedrós- Huber I (1997) Bacterial diversity on lake snow — phylo- Alió C (eds) Trends in microbial ecology. Spanish Society genetic and in situ analyses. PhD thesis, Technical Univer- for Microbiology, Barcelona, p 410–414 sity of Munich (in German) Berman T, Parparov A, Simon M (1999) Carbon fluxes in lim- Ikeda F, Shuto H, Saito T, Fukui T, Tomita K (1982) An extra- netic seston: relative significance of respiration, solubiliza- cellular polysaccharide produced by Zoogloea ramigera tion and non-photosynthetic formation. Arch Hydrobiol 115. Eur J Biochem 123:437–445 Spec Iss Adv Limnol 54:77–90 Jacobsen TR, Azam F (1984) Role of bacteria in fecal pellet Böckelmann U, Manz W, Neu TR, Szewzyk U (2000) Charac- decomposition: colonization, growth rates and mineraliza- terization of the microbial community of lotic organic tion. Bull Mar Sci 35:495–502 aggregates (‘river snow’) in the Elbe River of Germany by Kiørboe T, Lundsgaard C, Olesen M, Hansen JLS (1994) cultivation and molecular methods. FEMS Microbiol Ecol Aggregation and sedimentation processes during a spring 33:157–170 phytoplankton bloom: a field experiment to test coagula- Brachvogel T, Schweitzer B, Simon M (2001) Dynamics and tion theory. J Mar Res 52:297–323 bacterial colonization of microaggregates in a large Logan BE, Passow U, Alldredge AL, Grossart HP, Simon M mesotrophic lake. Aquat Microb Ecol (in press) (1995) Mass sedimentation of diatom blooms as large Brosius J, Dull TJ, Sleeter DD, Noller HF (1981) Gene organi- aggregates is driven by coagulation of transparent 150 Aquat Microb Ecol 25: 141–150, 2001

exopolymer particles (TEP). Deep-Sea Res 42:203–214 genetic diversity in a marine snow-associated bacterial Long RA, Azam F (1996) Abundant protein-containing parti- assemblage. Aquat Microb Ecol 14:261–269 cles in the sea. Aquat Microb Ecol 10:213–221 Reichenbach H (1992) The order Cytophagales. In: Balows A, Manz W, Amann R, Ludwig W, Wagner M, Schleifer KH Trüper HG, Dworkin M, Harder W, Schleifer KH (eds) The (1992) Phylogenetic oligonucleotide probes for the major prokaryotes, Springer, New York, p 3631–3675 subclasses of Proteobacteria: problems and solutions. Syst Schweitzer B, Huber I, Amann R, Ludwig W, Simon M (2001) Appl Microbiol 15:593–600 Alpha- and beta-proteobacteria control the consumption Manz W, Amann R, Ludwig W, Vancanneyt M, Schleifer KH and release of amino acids on lake snow aggregates. Appl (1996) Application of a suite of 16S rRNA-specific oligo- Environ Microbiol 67:632–645 nucleotide probes designed to investigate bacteria of the Shanks AL, Edmondson EW (1989) Laboratory-made artificial phylum cytophaga-flavobacter-bacteroides in the natural marine snow: a biological model of the real thing. Mar Biol environment. Microbiology 142:1097–1106 101:463–470 Mostajir B, Dolan JR, Rassoulzadegan R (1995) Seasonal vari- Shanks AL, Trent JD (1979) Marine snow: microscale nutrient ations of pico- and nano-detrital particles (DAPI yellow patches. Limnol Oceanogr 24:43–47 particles, DYP) in the Ligurian Sea (NW Mediterranean). Simon M, Bunte C, Schulz M, Weiss M, Wünsch C (1998) Aquat Microb Ecol 9:267–277 Bacterioplankton dynamics in Lake Constance Passow U, Alldredge AL (1994) Distribution, size and bacter- (Bodensee): substrate utilization, growth control, and ial colonization of transparent exopolymer particles (TEP) long-term trends. Arch Hydrobiol Spec Iss Adv Limnol in the ocean. Mar Ecol Prog Ser 113:185–198 53:195–221 Passow U, Alldredge AL, Logan BE (1994) The role of partic- Simon M, Grossart HP, Schweitzer B, Ploug H (in press) ulate carbohydrate exudates in the flocculation of diatom Microbial ecology of organic aggregates in aquatic blooms. Deep-Sea Res 41:335–357 ecosystems. Aquat Microb Ecol Pernthaler J, Glöckner FO, Unterholzner S, Alfreider A, Psen- Smetacek V (1985) Role of sinking in diatom-life history ner R, Amann R (1999) Seasonal community and popula- cycles: ecological, evolutionary and geological signifi- tion dynamics of pelagic Bacteria and Archaea in a high cance. Mar Biol 84:239–251 mountain lake. Appl Environ Microbiol 64:4299–4306 Smith DC, Simon M, Alldredge AL, Azam F (1992) Intense Phillips CJ, Smith Z, Embley TM, Prosser JI (1999) Phylo- hydrolytic enzyme activity on marine aggregates and genetic differences between particle-associated and implications for rapid particle dissolution. Nature 359: planktonic ammonia-oxidizing bacteria of the β-subdivi- 139–142 sion of the class Proteobacteria in the northwestern Snaidr J, Amann R, Huber I, Ludwig W, Schleifer KH (1997) Mediterranean Sea. Appl Environ Microbiol 65:779–786 Phylogenetic analysis and in situ identification of bacteria Ploug H, Grossart HP, Azam F, Jørgensen BB (1999) Photo- in activated sludge. Appl Environ Microbiol 63:2884–2896 synthesis, respiration and carbon turnover in sinking Weiss P, Schweitzer B, Amann R, Simon M (1996) Identifica- marine snow from surface waters of Southern California tion in situ and dynamics of bacteria on limnetic organic Bight: implications for the carbon cycling in the ocean. aggregates (lake snow). Appl Environ Microbiol 62: Mar Ecol Prog Ser 179:1–11 1998–2005 Porter KG, Feig YS (1980) The use of DAPI for identifying and Zwisler W (2000) Structural analysis of bacterioplankton counting aquatic microflora. Limnol Oceanogr 25:943–948 communities in Lake Constance. PhD thesis, University of Rath J, Wu KY, Herndl GJ, DeLong EF (1998) High phylo- Oldenburg (in German)

Editorial responsibility: Karel 2imek, Submitted: March 13, 2001; Accepted: June 28, 2001 >eské Budeˇjovice, Czech Republic Proofs received from author(s): August 22, 2001